Pyflink user defined aggregate function does not accumulate values and merge and accumulate methods do not seem to be working - merge

Question

I am running the following pyflink code, which accepts kafka messages and performs sliding window aggregations:
# Import modules
import os
from pyflink.datastream import StreamExecutionEnvironment
from pyflink.table import StreamTableEnvironment, EnvironmentSettings
from pyflink.table.window import Slide
from pyflink.table import DataTypes, TableDescriptor, Schema
from pyflink.table.expressions import lit, col, concat
from pyflink.table.table import Table
from pyflink.table.udf import ScalarFunction, udf
from pyflink.common import Row
from pyflink.table import AggregateFunction, DataTypes, TableEnvironment, EnvironmentSettings
from pyflink.table.expressions import call
from pyflink.table.udf import udaf
from pyflink.table.expressions import col, lit
from pyflink.table.window import Tumble
from pyflink.table.udf import udf
# Class to create alarm based on the number of ids in the window
class AlarmGen(AggregateFunction):
def create_accumulator(self):
return ['']
def get_value(self, accumulator):
num_ids = len(list(set(list(accumulator[0].split(',')[:-1]))))
if num_ids > 7:
return 'Alarm'
else:
return 'No Alarm'
def accumulate(self, accumulator, value):
accumulator[0] += value
def get_result_type(self):
return DataTypes.STRING()
def get_accumulator_type(self):
return DataTypes.STRING()
def merge(self, accumulator, accumulators):
accumulator += accumulators[0]
# Define environment
env = StreamExecutionEnvironment.get_execution_environment()
env.set_parallelism(1)
env_settings = EnvironmentSettings.new_instance().in_streaming_mode().build()
st_env = StreamTableEnvironment.create(env, environment_settings=env_settings)
# Define source table
st_env.execute_sql(
f"""
CREATE TABLE source (
id INT,
name STRING,
age INT,
ts BIGINT,
all_ids STRING,
rowtime as TO_TIMESTAMP(FROM_UNIXTIME(ts, 'yyyy-MM-dd HH:mm:ss')),
WATERMARK FOR rowtime AS rowtime - INTERVAL '0' SECOND
) WITH (
'connector' = 'kafka',
'topic' = '{os.environ["KAFKA_TOPIC"]}',
'scan.startup.mode' = 'latest-offset',
'properties.bootstrap.servers' = '{os.environ["KAFKA_HOST"]}',
'properties.zookeeper.connect' = '{os.environ["ZOOKEEPER_HOST"]}',
'properties.group.id' = '{os.environ["KAFKA_CONSUMER_GROUP"]}',
'format' = 'json'
)
"""
)
# Defie sink table
st_env.execute_sql(
"""
CREATE TABLE sink (
average_age DOUBLE,
window_end TIMESTAMP(3),
alarm STRING
) WITH (
'connector' = 'print',
'print-identifier' = 'Average Age, Window, alarm: '
)
"""
)
# Make functions available via udaf
alarm_gen = udaf(AlarmGen(), result_type=DataTypes.STRING())
# Execute function over a window
st_env.from_path("source")\
.window(Slide.over(lit(5).seconds)\
.every(lit(5).seconds).on("rowtime").alias("w"))\
.group_by("w")\
.select(col("age").avg, col("w").end, alarm_gen(col("all_ids")))\
.execute_insert("sink")\
.wait()
The inputs from kafka producer are in the following format:
{'id': 1, 'name': 'Osbourne', 'age': '73', 'ts': 1675667790, 'all_ids': '0,1,2,3,4,'}
The purpose of the AlarmGen class is to take the rows of the all_ids column, concatenate all the values together, then get all the unique values, and then to count them. If the count exceeds 7, the function should return 'Alarm', otherwise it returns 'No Alarm'.
Right now it seems the accumulator variable always remains as '' (an empty string), whether it is acting as an input into the accumulate function, or the merge function.
Then the accumulate function already receives '' as input, and so num_ids variable is always 0.
Any ideas why merge and accumulate methods may not be acting as they should be?
I am working with pyflink version 1.14.6.